The present invention is directed to probe stations adapted for making highly accurate low-current and low-voltage measurements of wafers and other electronic test devices. More particularly, the invention relates to such a probe station having a guarding system for preventing current leakage, a Kelvin connection system to eliminate voltage losses caused by line resistances, and an electromagnetic interference (EMI) shielding system.
The technique of guarding to minimize current leakage during low-current measurements, the use of Kelvin connections for low-voltage measurements, and the provision of EMI shielding are all well known and discussed extensively in the technical literature. See, for example, an article by William Knauer entitled “Fixturing for Low-Current/Low-Voltage Parametric Testing,” appearing in Evaluation Engineering, November, 1990, pages 150-153. See also Hewlett-Packard, “Application Note 356-HP 4142B Modular DC Source/Monitor Practical Application,” (1987) pages 1-4, and Hewlett-Packard, H-P Model 4284A Precision LCR Meter, Operation Manual (1991) pages 2-1, 6-9, and 6-15.
In guarding applications, a conductor surrounding or otherwise closely adjacent to a low-current line or circuit is maintained at the same potential as the line or circuit to reduce leakage currents therefrom, so that low-current measurements can be made accurately.
Kelvin connections compensate for voltage losses caused by line resistances which-would otherwise cause errors in low-voltage measurements. This is accomplished by providing a source line and a measurement line (also referred to commonly as “force” and “sense” lines, respectively) to an interconnection point (the Kelvin connection) which is as close to the test device as possible. A high-impedance voltmeter is connected to this interconnection point through the measurement line to accurately detect the voltage without any significant flow of current or resultant voltage drop in the measurement line. This avoids the error which would otherwise occur if the voltmeter were to detect the voltage through the source line, due to the voltage drop that occurs in the source line.
Probe stations have previously been used for conducting tests with guarding, Kelvin connection, and EMI shielding techniques. However the custom set-up of such probe stations required for guarding and Kelvin connection procedures is time-consuming and, in some instances, limited as to effectiveness. For example, in an article by Yousuke Yamamoto, entitled “A Compact Self-Shielding Prober for Accurate Measurement of On-Wafer Electron Devices,” appearing in IEEE Transactions on Instrumentation and Measurement, Volume 38, No. 6, December, 1989, pages 1088-1093, a probe station is shown having a respective detachable triaxial connector mounted on the probe card and the chuck assembly which supports the test device. The intermediate connector element of a triaxial connector normally is utilized for guarding. purposes. However the chuck assembly shown has only a chuck and a shield, with no separate guarding structure to which the intermediate connector element could be connected. Accordingly significant time-consuming alteration of such a station would be required to obtain both a guarded and shielded chuck assembly. The probes on the probe card, on the other hand, are both guarded and shielded; however there is no means of enabling each probe to be moved independently of the others in unison with its guard and shield to accommodate different contact patterns of test devices, thus sacrificing flexibility of the probe station. Also, there is no provision for Kelvin connections on the chuck assembly, which would require more than a single triaxial connector as shown.
Chuck assemblies are available which provide guarding and shielding components. For example, Temptronic Corporation of Newton, Mass. markets a thermal chuck assembly atop which is mounted an “add-on” supporting surface for the test device, with a copper guarding layer interposed between the add-on surface and the underlying chuck assembly and insulated from each by respective sheets of insulating material. This structure permits a signal line to be soldered to the add-on surface, a guard line to be soldered to the copper guarding layer, and a ground line to be soldered to the underlying chuck assembly which can then serve as a shield. However such wiring requires time-consuming set-up, particularly if Kelvin connections are also required. Moreover, the use of sheet insulation to insulate the copper guarding layer, from the add-on surface and the underlying chuck assembly fails to provide as low a dielectric constant between the respective elements as is desirable to minimize leakage currents in view of the low level of current to be measured.
With respect to -probe stations that are designed to accommodate the measurement of low levels of current, a sensitivity threshold is normally encountered below which further improvements in current sensitivity are difficult to reliably achieve. In most commercial probe stations that are of such design, this sensitivity threshold is typically reached at about 20-50 femtoamps. However, improvements in device fabrication and in the capabilities of commercially available test instrumentation make it desirable to reduce the sensitivity threshold to a level reliably within the single digit femtoamp range.
A particular difficulty encountered in low-level current measurements is the excessive time required for measurement voltages to stabilize with reference to the device under test after a shift in voltage has occurred at the electrical input to the probe station. This problem of excessive settling time, as it is referred to, increases as the level of current under measurement is reduced. That is, due to the residual capacitance existing between spaced apart conductors in the region surrounding the immediate test site, a certain amount of time is required for the conductors that are in direct connection with the test device to fully charge or discharge to their desired voltages, and the time required will increase as the current through the device decreases. If the residual capacitance and the degree of input voltage shift are moderately large and if the level of current being measured is moderately small, the probe station operator can encounter settling times that are upwards of two or three minutes. Clearly, then, it is desirable that settling times be generally reduced in order to reduce overall measurement time, particularly where the device under test is a wafer containing large numbers of discrete devices, each of which may individually require low-level current testing.
In addition to settling effects, measurements of low level currents are also susceptible to error due to electrical discharge effects which occur because of the acceptance and release of charge by nonconductors in the region surrounding the immediate test site. At very low currents, these discharge effects can significantly-distort measurement values and hence contribute to unacceptable levels of measurement instability.